Another, simpler (I think) way to look at this is from an impedance-matching perspective:
For maximum power transfer between any amplifier circuit and its load, the output impedance of the amp needs to match the speaker.
Noooooo!!!11
This stupid thing, taught in university classrooms, goes around and around misunderstood from decade to decade.
Well obviously, from a very limited perspective, it is true. If you have an arbitrary Thevenin equivalent source, say an amplifier which output transistors define output impedance of 10 ohms, then if you do want to get the maximum possible power from it, you match the output load of 10 ohms.
But then you make the logical mistake and turn it around - if you have a 10 ohm load, to get maximum power possible, do you use a 10 ohm output impedance amplifier? Obviously, no! Use 1 ohm, and you'll get
even more power. Run the math, it's trivial!
The key you didn't think about in school is this:
Often:
* the power used by the load does actual work (is converted to sound in a speaker, is converted to mechanical energy in a motor, is converted to light in a lightbulb...)
* the power used by the amplifier is wasted power (converted to heat in the amplifier)
From this perspective, matching the impedances so that power use is shared 50%-50% is a ridiculous idea. Instead, if you choose Zamplifier = 0 and Zload = whatever to bring the voltage & current relationship to the most optimal point for practical design, you have just the amount of power you
need, with perfect load voltage regulation, and best possible amplifier efficiency.
The imaginary amplifier with 10 ohm output impedance was probably never meant to drive a 10 ohm load, but maybe a 1000 ohm load. Why?
Because amplifiers are often designed to be voltage sources that can drive the output to whatever value within its rails. If it had the equal output and load impedance, then the output voltage would change by factor of 50%. Try to output 10 volts - only 5 volts available to the load. 5 volts wasted as heat in the amplifier (this obviously won't matter in a linear amplifier which wastes power as heat anyway).
This is why amplifiers have the feedback, which is designed to bring the output impedance close to zero. The transistors and PCB traces themselves still have some on-resistance, so the feedback needs some voltage leeway to work with. Want to output exact 10V, to a load which hogs 1A? Say, transistors drop 1V because of their internal on-resistance. Is the output dropped to just 9V, and is the output impedance 1V/1A = 1 ohm? No, and no. The feedback measures the output and adjusts the transistors to be "more on" until the output is at 10V, and the output impedance is hence 0V/1A = 0 ohms.
So whenever you want to drive a load with a voltage source, the correct impedance is not the load impedance, but zero. In case of linear amplifiers driving intermediate levels, it's about voltage regulation, giving the load the correct voltage. In case of switching amplifiers, it's all about heat as well. If you match the source and load impedance, the efficiency of the amplifier is limited to just 50%, which would be very crappy for such circuit.